Apart from the randomForest package, the party package also provides an implementation of random forest. In the following steps, we illustrate how to use the cforest function within the party package to perform classifications:
- First, install and load the party package:
> install.packages("party") > library(party)
- You can then use the cforest function to fit the classification model:
> churn.cforest = cforest(churn ~ ., data = trainset, con-
trols=cforest_unbiased(ntree=1000, mtry=5))
> churn.cforest
Output
Random Forest using Conditional Inference Trees
Number of trees: 1000
Response: churn
Inputs: international_plan, voice_mail_plan, num-
ber_vmail_messages, total_day_minutes, total_day_calls, to-
tal_day_charge, total_eve_minutes,
total_eve_calls, to-tal_eve_charge, total_night_minutes,
total_night_calls, to-tal_night_charge, total_intl_minutes,
total_intl_calls, to-tal_intl_charge, number_customer_service_calls
Number of observations: 2315
- You can make predictions based on the built model and the testing dataset:
> churn.cforest.prediction = predict(churn.cforest, testset,
OOB=TRUE, type = "response")
- Finally, obtain the classification table from the predicted labels and the labels of the testing dataset:
> table(churn.cforest.prediction, testset$churn) Output churn.cforest.prediction yes no yes 91 3 no 50 874